The reasonable station layout and scheduling optimization could improve the efficiency of the public bicycle system effectively. It is of great importance to analyze and forecast the spatial and temporal distribution of demand. Therefore, based on the rental data of the public bicycle system of Hohhot in 2017, this study establishes a public bicycle demand forecasting model by using Back Propagation (BP) neural network. Firstly, the station with a large demand is selected to analyze the distribution of demand in different periods. By analyzing the similarity of data in different periods, the law of data variation is found out. Then the public bicycle demand forecasting model based on BP neural network (PBDF-BP model) is trained and validated by using the peak hour rental data. Moreover, by comparing the prediction accuracy with other forecasting models, it is found that the PBDF-BP model has less error in prediction results and higher stability. The findings in this study may help to promote the sustainable development of the public bicycle system.

Demand forecasting;Public Bicycle;BP Network